基于SPIHT算法的高光谱图像压缩研究

发布时间：2018-06-15 10:54

本文选题：高光谱图像 + KLT变换　；参考：《成都理工大学》2016年硕士论文

【摘要】：高光谱遥感图像具有光谱分辨率高、谱段窄等特点,在为我们提供更为丰富的地物信的同时,它的数据量也变得十分庞大,为其传输或存储都带来了不便,限制了它在实际中的应用,所以研究高效的压缩方法是十分有必要的。本文对高光谱图像的压缩技术作了研究,分析了高光谱图像的两种冗余,本文采用KLT变换去除谱间冗余,小波变换去除空间冗余,对变换后的系数采用两种SPHIT编码方案:一是使用二维SPHIT算法对各个波段进行编码;二是采用3D-SPHIT算法对整体进行编码;在二维SPHIT编码中,采用了基于KLT变换特征向量的非均匀码率分配方法,实验结果表明,该方法获得的PSNR要比均匀分配码率的方法高出约3db。在3D-SPHIT编码中,分析了KLT变换结合小波变换与直接采用三维非对称小波变换对高光谱图像冗余去除的性能,结果表明,总的来说,采用KLT变换结合小波变换的方案获得的SNR要高于采用三维小波变换的方案,在低码率时,两者相差不大,但随着码率的增加,采用KLT变换结合小波变换的方案可以获得更好的SNR。
[Abstract]:Hyperspectral remote sensing images have the characteristics of high spectral resolution and narrow spectral bands. While providing us with more abundant information of ground objects, the amount of data in hyperspectral remote sensing images has become very large, which brings inconvenience to its transmission or storage. It limits its application in practice, so it is necessary to study the efficient compression method. In this paper, the compression technology of hyperspectral image is studied, and two kinds of redundancy of hyperspectral image are analyzed. In this paper, the inter-spectral redundancy is removed by KLT transform, and the spatial redundancy is removed by wavelet transform. Two kinds of SPHIT coding schemes are used for the transformed coefficients: one is to encode each band by using two-dimensional SPHIT algorithm, the other is to code the whole by using 3D-SPHIT algorithm. The non-uniform bit rate allocation method based on KLT transform eigenvector is used. The experimental results show that the PSNR obtained by this method is about 3 db. higher than that of the uniform allocation method. In the 3D-SPHIT coding, the performance of combining KLT transform with wavelet transform and directly using 3D asymmetric wavelet transform to remove redundancy of hyperspectral image is analyzed. The results show that, The SNR obtained by using KLT combined with wavelet transform is higher than that with 3D wavelet transform. At low bit rate, there is no difference between the two schemes, but with the increase of code rate, a better SNR can be obtained by using KLT combined with wavelet transform.
【学位授予单位】：成都理工大学
【学位级别】：硕士
【学位授予年份】：2016
【分类号】：TP751

【参考文献】